Diffusion-based Image Translation using Disentangled Style and Content Representation

TL;DR

This paper introduces a diffusion-based unsupervised image translation method that leverages disentangled style and content representations, improving content preservation and style transfer flexibility through novel loss functions and strategies.

Contribution

The paper proposes a new diffusion model approach using disentangled style and content features, including a semantic divergence loss and resampling, to enhance image translation quality.

Findings

Outperforms state-of-the-art models in text-guided translation

Maintains better content preservation during diffusion

Effective style transfer guided by CLIP and ViT features

Abstract

Diffusion-based image translation guided by semantic texts or a single target image has enabled flexible style transfer which is not limited to the specific domains. Unfortunately, due to the stochastic nature of diffusion models, it is often difficult to maintain the original content of the image during the reverse diffusion. To address this, here we present a novel diffusion-based unsupervised image translation method using disentangled style and content representation. Specifically, inspired by the splicing Vision Transformer, we extract intermediate keys of multihead self attention layer from ViT model and used them as the content preservation loss. Then, an image guided style transfer is performed by matching the [CLS] classification token from the denoised samples and target image, whereas additional CLIP loss is used for the text-driven style transfer. To further accelerate the semantic change during the reverse diffusion, we also propose a novel semantic divergence loss and resampling strategy. Our experimental results show that the proposed method outperforms state-of-the-art baseline models in both text-guided and image-guided translation tasks.